2014 Play-Action Offense
by Sterling Xie
(Ed. Note: With this article, you can welcome Sterling Xie to the FO family as an official staff writer. He's been interning for a few months, and he's also written in the past for Bleacher Report and Advanced Football Analytics. I told Brian Burke FO was going to steal him away, but now that Brian has moved to ESPN full-time, we can just legally adopt him instead. By the way, Sterling is a Patriots fan, just like our other recent add, Andrew Healy, so your accusations of Patriots bias from Football Outsiders are now THREE TIMES MORE VALID! Very exciting, I'm sure. -- Aaron Schatz)
As we churn away at Football Outsiders Almanac 2015, we'll start reviewing results from our 2014 charting data, much of which will appear in the book. Play-action passing is a fun annual staple to think about, and the stats on it provide valuable but also wildly inconsistent information. Our information on which plays involve a play-action fake comes courtesy of ESPN Stats & Information.
League-wide, the rate of play-action usage held steady from the 2013 season at 21 percent. The distribution evened out slightly, as the standard deviation in play-action rate fell to 5.58 percent after sitting at 6.09 percent in 2013. Two outliers skewed the 2014 figure (don't worry, we'll get to them shortly). Play-action success went up slightly, as the league average DVOA jumped from 20.6% to 24.0% last fall. However, the DVOA on non-play-action passes also rose from 9.6% to 12.3%, so increased efficacy certainly wasn't unique to play-action. In general, aerial attacks are simply strengthening their vise grips over ground games.
A couple changes to this year's article: We aren't showing play-action rates by down, because they have essentially remained constant for five years, particularly on second and third down. On the big table below, we're also not fleshing out the details by listing both play-action DVOA that includes scrambles and play-action DVOA that doesn't include scrambles, because you have probably fallen asleep reading this sentence. This year's play-action DVOA listed below always includes both passes and scrambles. The only team with a DVOA difference greater than 4% was the New York Jets, whose play-action DVOA tumbled from 6.9% to -1.6% when excluding quarterback runs. Despite their reputations, Geno Smith and Michael Vick weren't actually good scrambling quarterbacks -- the former compiled -65 rushing DYAR, worst among all signal-callers -- so this is another indictment of Gang Green's miserable passing situation.
The following data is only for the 2014 regular season. Offenses are sorted by descending rate of play-action usage (percentage of dropbacks). The first two splits compare results on play-action passes versus passes without play-action. The final "Difference" section shows the change in yards per play and DVOA when play-action isn't used. A high ranking there indicates the offense was much better with play-action.
| With PA (Pass/Scramble) | No PA | Difference | |||||||||
| Offense | PA Pct | Rank | Yds/Play | DVOA | Rank | Yds/Play | DVOA | Rank | Yds/Play | DVOA | Rank |
| PHI | 33% | 1 | 8.1 | 41.3% | 6 | 6.2 | -2.4% | 24 | 1.9 | 43.7% | 3 |
| CLE | 31% | 2 | 8.8 | 32.7% | 14 | 5.4 | -6.9% | 27 | 3.4 | 39.6% | 5 |
| SEA | 31% | 3 | 7.6 | 31.5% | 15 | 6.7 | 29.8% | 7 | 0.9 | 1.6% | 22 |
| KC | 30% | 4 | 7.7 | 39.5% | 7 | 5.2 | 0.6% | 22 | 2.5 | 38.9% | 6 |
| CAR | 28% | 5 | 6.3 | 22.5% | 20 | 6.0 | 4.1% | 21 | 0.2 | 18.4% | 13 |
| MIA | 26% | 6 | 6.0 | -4.0% | 28 | 5.9 | 27.9% | 8 | 0.1 | -31.8% | 29 |
| NE | 25% | 7 | 8.3 | 46.8% | 5 | 6.1 | 31.5% | 6 | 2.2 | 15.3% | 16 |
| GB | 24% | 8 | 9.3 | 60.0% | 2 | 7.3 | 47.9% | 1 | 2.0 | 12.2% | 18 |
| JAC | 23% | 9 | 5.5 | -15.4% | 29 | 4.9 | -32.2% | 32 | 0.6 | 16.8% | 15 |
| NYG | 23% | 10 | 7.8 | 35.4% | 11 | 6.6 | 11.3% | 16 | 1.2 | 24.1% | 9 |
| MIN | 23% | 11 | 5.4 | -28.6% | 31 | 6.0 | 4.9% | 20 | -0.6 | -33.6% | 30 |
| CIN | 23% | 12 | 7.1 | -25.3% | 30 | 6.3 | 16.7% | 14 | 0.9 | -42.1% | 31 |
| DEN | 23% | 13 | 10.3 | 33.9% | 13 | 6.8 | 45.0% | 3 | 3.6 | -11.2% | 26 |
| WAS | 22% | 14 | 10.2 | 55.1% | 4 | 5.7 | -23.0% | 31 | 4.5 | 78.2% | 1 |
| PIT | 22% | 15 | 8.8 | 58.9% | 3 | 7.4 | 46.5% | 2 | 1.3 | 12.4% | 17 |
| OAK | 21% | 16 | 5.0 | 25.2% | 19 | 5.1 | -15.1% | 30 | -0.1 | 40.3% | 4 |
| With PA (Pass/Scramble) | No PA | Difference | |||||||||
| Offense | PA Pct | Rank | Yds/Play | DVOA | Rank | Yds/Play | DVOA | Rank | Yds/Play | DVOA | Rank |
| NYJ | 21% | 17 | 7.1 | 6.9% | 24 | 5.1 | -5.2% | 25 | 2.0 | 12.1% | 19 |
| CHI | 21% | 18 | 6.5 | 34.9% | 12 | 6.0 | 10.7% | 17 | 0.5 | 24.2% | 8 |
| SF | 20% | 19 | 5.3 | -1.4% | 27 | 6.5 | 18.7% | 12 | -1.2 | -20.1% | 28 |
| HOU | 19% | 20 | 7.9 | 30.8% | 16 | 6.3 | 10.3% | 18 | 1.6 | 20.5% | 11 |
| DET | 19% | 21 | 6.7 | 5.2% | 25 | 6.2 | 12.0% | 15 | 0.5 | -6.8% | 25 |
| IND | 19% | 22 | 7.9 | 38.9% | 8 | 7.1 | 16.8% | 13 | 0.8 | 22.1% | 10 |
| BAL | 18% | 23 | 8.5 | 30.3% | 17 | 6.7 | 33.4% | 5 | 1.8 | -3.1% | 24 |
| STL | 18% | 24 | 7.2 | 12.9% | 22 | 6.0 | -6.7% | 26 | 1.2 | 19.7% | 12 |
| NO | 18% | 25 | 7.4 | 37.9% | 10 | 7.0 | 20.1% | 11 | 0.4 | 17.8% | 14 |
| ATL | 17% | 26 | 7.4 | 15.1% | 21 | 6.8 | 27.4% | 9 | 0.6 | -12.3% | 27 |
| ARI | 17% | 27 | 6.7 | 9.9% | 23 | 6.6 | 7.6% | 19 | 0.1 | 2.3% | 21 |
| DAL | 17% | 28 | 9.7 | 38.6% | 9 | 7.3 | 39.1% | 4 | 2.4 | -0.5% | 23 |
| TEN | 16% | 29 | 8.6 | 25.4% | 18 | 5.7 | -11.4% | 29 | 2.9 | 36.8% | 7 |
| BUF | 15% | 30 | 5.4 | 4.4% | 26 | 5.9 | 0.4% | 23 | -0.5 | 4.0% | 20 |
| TB | 11% | 31 | 4.1 | -55.9% | 32 | 6.1 | -9.1% | 28 | -1.9 | -46.8% | 32 |
| SD | 8% | 32 | 9.6 | 90.3% | 1 | 6.6 | 26.7% | 10 | 3.1 | 63.6% | 2 |
| NFL | 21% | 7.5 | 24.0% | 6.3 | 12.3% | 1.2 | 11.7% | ||||
Comments
26 comments, Last at 24 Jun 2015, 7:43pm
#7 by Alternator // Jun 20, 2015 - 12:58am
The Plexiglass Principal says that any big, sudden change from one year to the next is likely to see a smaller, rebound change giving back a portion of those gains.
Basically, to suddenly improve (or decline) dramatically, it's likely that there was a huge dose of good (or bad) luck involved. Since luck is, by definition, likely to balance out and end up closer to average, the level of performance is likely to regress (or rebound).
Example:
The Vegas Bookies go from 3-13 in 2017 and 2-14 in 2018, to 10-6 in 2019. For that big a swing, the smart money says that it's a bad team that improved some, but also got a ton of lucky breaks, and their true talent level is probably more like 7-9. When the luck vanishes the next year, hey, they end up 8-8. That's the Plexiglass Principal.
#3 by Cleared for Contact // Jun 19, 2015 - 2:39pm
The Charger paltry play action doesn't seem too surprising when you look at their rushing offense:
30th overall
25th in DVOA
3.4 yards a carry
They probably only thought to use play action on the rare occasions they run game was effective enough to force the defense's hand.
Does shotgun effect this too? They have used it a lot more since McCoy took over.
#5 by Vincent Verhei // Jun 19, 2015 - 7:34pm
Philadelphia used shotgun on 86 percent of all snaps last year, more than anyone in the league, and they still led the NFL in play-action rate, so no, that doesn't have anything to do with it. (San Diego was third in shotgun usage behind PHI and Miami.)
#6 by Sterling Xie // Jun 20, 2015 - 12:32am
There also hasn't been a ton of proof that run game success leads to better play-action, though I have no clue if coaches think that way regardless. But SD is the best example of how a poor running game shouldn't really affect how good their play-action is.
#9 by Thomas_beardown // Jun 20, 2015 - 4:45pm
The Eagles had the highest rate and they still ranked 6th.
So, even if they fell from 90% DVOA to idk 50% if they went to most used, that would be well worth it because it would still be far better than their normal offense.
#10 by Sterling Xie // Jun 20, 2015 - 9:49pm
Agreed, but I think it's unlikely that 8 % usage was the optimal rate. And remember, SD was 10th in non-PA DVOA. This was just a straight-up elite passing offense, and even if they had used it around 30 % like the league leaders, I doubt it would've gone poorly. Probably rhetorical anyways, since Philip Rivers has clearly gotten very comfortable in a system that last 2 years that doesn't include much PA.
#11 by jtr // Jun 22, 2015 - 9:00am
You mention in the article that PA passing DVOA does not correlate with itself well year-to-year. Has a correlation ever been done between year N's PA DVOA and year (N-1)'s overall passing DVOA? That might correlate better simply by having a larger sample of the previous season's passing ability.
#12 by OSS117 // Jun 22, 2015 - 9:52am
Play-action invokes images of the QB turning his back to the D, faking the hand-off then squaring up off a deep drop to launch an ICBM over a sucked up D. Or boot-action off a stretch to do the same. Notsomuch a face forward and ponied QB holding the ball like a candle vigil as his sidecar passes in front of him before flicking it out to a TE screen.
I'd think read/option would be worth a huge mention in this piece. Since they populate the top of the board. And I don't know if I really consider read/option a true play-fake/play-action. And it's more or a package play really than read/option because in most cases the DE isn't being 'read' and is being blocked. The 'read' is being made presnap, and not determined by the actions of the DE. It's just two plays in one that look and execute exactly the same, except for the decision of the QB. The play fake isn't really selling anything. The defense is already committed long before the play fake. Maybe they're more successful because they're high percentage plays, because inherent in the design is that the defense is always wrong, rather than anything the playfake might accomplish.
Just saying, hard to really compare a read/option face-forward pony/gun play-action with the more traditional under-center type play-action. Very, very different.
Roethlisberger when Arians took over was good with play-action. On those rare occasions when the protection actually held up. But he was getting killed more often than not because the pressure was on him before he could even turn and square. So they ended up pretty much scrapping it. Because it's 7 step drops or nothing in Brucey's O. And Roethlisberger was already getting plenty killed enough without the added burden of play-action. I'm guessing it was the same for Rivers last year. They might use more shotgun, but from what I recall seeing, the play-action was with Rivers under center.
#13 by runaway robot // Jun 22, 2015 - 12:16pm
Great post! As far as Arians and Roethlisberger, it goes a long way toward explaining the mysterious absence of what seemed to have been an excellent play action game in Ben R.'s first couple years.
P.S. How do you feel about Lost in Rio? Would you agree that Cairo: Nest of Spies was better?
#16 by Sterling Xie // Jun 22, 2015 - 3:41pm
I agree that they're fundamentally different types of plays even though both true play-action and read-option play-action get lumped together. But we also don't have a large enough sample to see how these read fakes differently affect the defense compared to traditional play-action. I think it's interesting that Philly was still a really good PA team even though their passing game as a whole declined big time (as you can tell by the non PA data). If packaged plays continue to make up a larger and larger fraction of "play-action" passes, we'll see how that affects the overall numbers.
#19 by OSS117 // Jun 23, 2015 - 9:29am
It is interesting. Do you have completion percentages, reception length, yards per completion, and YAC breakouts/splits for with and without play-action? If I had to guess, I'd say the non-PAs the pass attempts are going more downfield at a much reduced completion percentage.
#17 by dank067 // Jun 22, 2015 - 11:11pm
Agree with you that read option offenses might warrant a closer look going forward. But packaged plays don't necessitate a play fake. Many teams (often ones that don't use option tactics) run package plays where, for example, the QB aborts the run scheme right at the snap or after two steps or what have you and makes a quick throw. So even when the play is predetermined, why do zone read teams still bother to show the fake? They obviously see some value in it, and I think it goes back to the root of what you're trying to accomplish with play action in the first place: trying to make the offense less predictable so the defense can't anticipate, cheat, etc.
#18 by OSS117 // Jun 23, 2015 - 9:06am
Apologies if I was unclear. I'm not saying all package plays are read/option. Or that all package plays have a play-action element. I am talking about read/option offenses in the NFL specific to play-action. And why I think it's worth mentioning.
And I'm not even talking about all read/option in the NFL. Just general observations, and that's that the QB isn't really reading the DE post snap, rather the entire D and determining presnap where it is disadvantaged. Like with package plays. The routineness of the play requires execution to be carried out the same, regardless of the option chosen. Nobody should really know what the option is but the QB. The offensive line blocks the same, the receivers all carry out the same routes, the back prepares for a handoff that may or may not come, and hits the hole the same, etc. And they are usually quick hitting plays. Because you can't have your OL downfield blocking for a run while the QB is surveying the field looking to unload it deep. The protection isn't there for it to be anything but a quick hitting play. If it's a pass the ball has to come out fast. On these read/option play-action plays, or package pass plays in general, the ball travels maybe +/- 3 yards of the LOS. And released well under 2 seconds of the snap. Polar opposite of traditional under-center play-action. The only thing they have in common is the play-fake. And IMO the play-fake in read/option or with package plays have very little to do with the results of the play.
Because the ball is going where the D is disadvantaged and because they are quick plays where the D doesn't really have a chance to react one way or the other to a play-fake, IMO the success of these plays has much, much more to do with the inherent high percentage success of the play design and decision of the QB than it has to do with the act of play-action. That's why package plays are becoming the bread and butter plays of most offenses now. 1st and 2nd down plays, maybe little big play potential, but good for 5-10 quick cheap yards and keeping the O on schedule and the chains moving.
And that's why I'm suggesting this is an apples and oranges list. You can't really make many comparisions or determinations from this list. You can't say 'Look at what Philly did, it's proof that SD should've been more confident with their play-action." Or that their play-action could've maintained under greater utilization. I think the data could benefit from refining/break-out to either QB depth or Under-center vs Shotgun/Pony. Yeah the samples would be much smaller, but I think it would better serve context and comparisons. Or at the least identify how each team is predominantly utilizing play-action. I'm not looking to be critical. I just think it's a very important point/distinction absent in what is otherwise a very interesting and well-written piece.
#20 by dank067 // Jun 23, 2015 - 10:13am
I hear you, and I definitely think it's worth further investigation. It's just that my first thought was that, even though the Eagles are often using package plays with predetermined reads by the QB, they're still showing the fake at the mesh point when they seemingly don't have to—why? Otherwise you could maybe get the ball out even another second faster.
As you and Sterling mentioned, the fakes themselves might be not affecting the defense or affecting it in a different way (or just reflect a very different goal by the offense), in which case this is definitely apples vs. oranges as you suggest. I just saw in the fact that they value the play fake a kind of unifying point with "standard" play action. But yeah, I would also be interested in studying it to see how the data comes out.
#21 by Thomas_beardown // Jun 23, 2015 - 10:32am
Regarding the last paragraph, even if SD's playaction took a severe efficiency hit from higher utilization, up to being only half as effective, it would have been worth it because that still would have been more effective than their non-pa offense.
So regardless of what the Eagles are doing, the Chargers should use PA more.
#25 by OSS117 // Jun 24, 2015 - 10:24am
There may be a reason why they shouldn't. But on paper, yeah, looks that way. I read something years ago by Brian Burke about imbalanced play calling, as it related to down/distance and run v pass. That you chose the option with the better payoff until the defense adjusts and it reaches equilibrium. I suppose in theory the same could be applied/refined here with regards to play-action v no play-action.
#22 by Vincent Verhei // Jun 23, 2015 - 2:21pm
Because some have asked for it:
All teams, play-action, under center vs. shotgun, 2014:
Shotgun: 7.2 yards per play, 7.5 avg. pass distance, 53% success rate, 23.0% DVOA
Under center: 7.7 yards per play, 11.0 avg. pass distance, 51% success rate, 25.0% DVOA
That's a big difference in average throw, but the pros and cons of that largely balance out in DVOA.
#26 by Bobman // Jun 24, 2015 - 7:43pm
Interesting that Colts were decent, given that Richardson was their most-used RB in 2014 (and 2013). Unless they got a boost because the defenders were laughing so hard whenever Luck faked to #34. "Andrew, man, you kill me; that is so damn funny when you... oh crud."
I assume Luck will improve in 2015 given that Gore is an actual threat.